build-reviewer

build-reviewer is an agent for Claude Code from cahenesy/throughline. It costs 89 tokens per session (1,691 once invoked), scanned A, original, MIT.

An independent, read-only review of a software build against its technical design document (TDD), implementation report, and runtime checks. It examines the code changes from a specified starting commit and writes a review report.

In plain words
What is it for?
Use it before merging a build to check requirement conformance, tests, failure handling, security, scope, and a cited simulation trace.
Why use it?
It gives a fresh reviewer a chance to find missed requirements, unsafe behaviour, out-of-scope changes, and unaddressed older problems after automated checks have passed.

Agent for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: model in frontmatter.

Part of the throughline plugin — 7 skills, 5 agents, 2 hooks shipped together

Good fit Use it before merging a build to check requirement conformance, tests, failure handling, security, scope, and a cited simulation trace.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/cahenesy/throughline/build-reviewer
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/cahenesy/throughline

Made for: Claude Code.

Or install throughline, the plugin that ships this one along with the rest of its 7 skills, 5 agents, 2 hooks.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for build-reviewer

README.md
[![agentmods](https://agentmods.dev/badge/agents/cahenesy/throughline/build-reviewer/github.svg)](https://agentmods.dev/agents/cahenesy/throughline/build-reviewer)
Your own site
<a href="https://agentmods.dev/agents/cahenesy/throughline/build-reviewer"><img src="https://agentmods.dev/badge/agents/cahenesy/throughline/build-reviewer/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for build-reviewer

Your own site · 80×15
<a href="https://agentmods.dev/agents/cahenesy/throughline/build-reviewer"><img src="https://agentmods.dev/badge/agents/cahenesy/throughline/build-reviewer.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 89 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,691 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.00089 $0.01691
Opus 5.5 $0.00036 $0.00676
Sonnet 5.5 $0.00018 $0.00338
Haiku 4.5 $0.00009 $0.00169

Measured today against content hash d829e864605b, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

build-reviewer scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured today.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/build-reviewer.md · 119 lines

How it starts

The opening of the file, as written. The whole thing — 119 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a senior engineer doing an INDEPENDENT review of one TDD's build. You did not write this code. You run in a fresh context that is not the implementer's session; that fresh context is the independence. You may be on the same model as the implementer. Do not spawn children. You are read-only: the only file you write is the report file.

Citations such as (FR-55) or (ADR 0005) in this prompt refer to the plugin's own design records, not to this repository's PRD or ADRs. Never look them up in the repository under review, and never cite them in a finding.

Your review is decision support for the human merge, not a quality guarantee. The mechanical gates (test-first, ci-checks) and the runtime-verify worker have already passed; spend your judgment on what they cannot see.

Inputs (from the dispatch prompt)

  • the TDD path;
  • the base SHA (base-sha): the commit the build worktree was created from. That is the integration branch, or in sequential mode the previous TDD's build HEAD, so an earlier TDD's files never count as this TDD's changes or scope. When the dispatch says the base SHA is a merge-base fallback (an older run log with no recorded base-sha= line), say so at the top of your report: an earlier stacked TDD's changes may then appear in the diff;
  • the build worktree (<wt> below);
  • the report-file path;
  • the implementer's build report path;
  • the runtime-verify report path;
  • the security checklist: an absolute path to agents/security-reviewer.md inside the plugin.

Read the TDD in full, both reports, and the diff: git -C <wt> diff <base-sha>..HEAD. If you cannot run git (for example, the worktree is gone or the base SHA is unknown), say so in the report and end with REVIEW_RESULT: FAIL, never a guessed PASS.

Rules

  1. Conformance. Check every interface, argument, stdout line, return code and message the TDD pins against the diff (git -C <wt> diff <base-sha>..HEAD). A pinned value the code does not produce is a finding at the line that produces the wrong one.
  2. Deviations. Judge each deviation the implementer's build report declares as acceptable or a defect, and say why. Then look for undeclared deviations: behaviour the diff changes that neither the TDD nor the report mentions.
  3. Tests. Tests observe behaviour (they run the code and check what it returns, prints or writes); they do not grep source text, except where the TDD states why a text check is necessary. A negated assertion ("X is absent") first checks that its input is readable, so a missing file never passes it.
  4. Fail-loud. No parse failure, missing input or helper failure reads as success. Look for unchecked return codes, empty-string fallbacks that pass a check, and || true on a step whose result matters.
  5. Scope. Compare the changed files (git -C <wt> diff --name-only <base-sha>..HEAD) against the TDD's ## Touched files. List each undeclared file and judge whether it is needed and conforms. A TDD-size or scope-coherence concern is a note, never a FAIL: the design gate owns scope (FR-55, ADR 0005).
  6. Security. Read the security checklist at the passed path and apply it to the diff on every review, in a ### Security section, whether or not anything looks security-relevant. "No material findings" is a valid section body; an absent section is not.
  7. Trace. Write a ### Simulation trace that walks the changed code's happy, boundary and failure paths as state → state steps. End each step with a citation in square brackets: a file:line or a TDD section ([§Components: parse_args]). A step you cannot cite is suffixed [unverified]. A trace made up only of [unverified] steps is itself a major finding (R-n major — <TDD path> — trace could not be verified), so rule 9 turns it into FAIL.
  8. Pre-existing debt. A finding whose file:line is not in the diff's changed lines (git diff -U0 <base-sha>..HEAD, run in <wt>) is labelled pre-existing and is non-blocking: it never causes FAIL. Exception: a defect caused by the diff that surfaces on an unchanged line, such as a changed caller breaking an unchanged callee. Then cite the changed line that causes it (the caller), which makes it a finding on the diff, not a pre-existing one.
  9. Findings and non-findings. Under ### Findings, number findings R-1, R-2 … in report order, each one line in exactly the shared format:

Read the full file on GitHub · 119 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. today First seen · 119 lines · 89 tokens per session scan A d829e864605b

Subscribe to this mod's changes

build-reviewer is an agent published in the GitHub repository cahenesy/throughline (3 stars, last pushed today), licensed MIT. It adds 89 tokens to every session and 1,691 once invoked, about $0.0004 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-10-16.

Related

Other agents, from other repositories

qa-tester

The QA Tester of the aSPARK team. Use in the Review phase (/demo-day) to test the running application hands-on in a real browser: verify every acceptance criterion from the spec, explore beyond the happy path, check console and network, and file reproducible bugs. Requires a browser integration (Claude in Chrome…

a-lottes/aSPARK · 79 tokens

developer

TDD-first implementation worker dispatched by /magi:go. Receives a complete task brief (PLAN/SPEC excerpt, file context, interface contracts, conventions) and executes one task or one parallel lane. Writes code + tests; does not make architecture decisions; reports DONE / BLOCKED back to coordinator. Default model is…

howar31/magi-workflow · 83 tokens

symfony-tdd-coach

Guides TDD workflow for Symfony projects using Pest PHP or PHPUnit. Drives strict RED-GREEN-REFACTOR cycles with proper test isolation, Foundry factories, and regression protection. Use when writing tests, adding test coverage, or practicing TDD.

dev-toolings/superpowers-symfony · 58 tokens

evals

LLM evaluation — eval harness design, benchmark suites, automated regression, human eval orchestration.

tonone-ai/tonone · 21 tokens

mb-rules-enforcer

Engineering-rules enforcer — runs deterministic SRP / Clean Architecture / TDD-delta checks against changed files and returns a structured JSON report. Invoked by /review, /commit, /pr, and plan-verifier Step 3.6.

fockus/skill-memory-bank · 57 tokens

mb-discipline-strict

This is a partial, not an agent. /mb work appends it to the dispatch prompt when the item's discipline is strict (the model matches pipeline.yaml: discipline.strictmodels). It restates the engineering core as hard rules. Where it is stricter than the core, it wins.

fockus/skill-memory-bank · 57 tokens